Max Topchii -

How data analytics drives wind turbine power generation

Data analytics and related software lies behind the burgeoning wind turbine industry, with predictive maintenance to the fore. We find out how

The UK used to rely on the oil and gas that lay beneath the North Sea. Now it increasingly harnesses the winds that blow over it, accelerating the use of data and specialist software to plan, operate and maintain the country’s rapidly growing fleet of offshore wind turbines.

In 2020, government figures showed that the UK obtained 4% of its power from the wind, 10 times as much as in 2010, as part of efforts to cut carbon emissions by shifting to renewable energy. Just over half was generated offshore, with that year seeing the opening of Hornsea One, the world’s largest offshore wind farm, consisting of 174 giant turbines 75 miles off the Yorkshire coast.

Its Danish operator, Ørsted, plans to open the even-bigger Hornsea Two this year, has consent to build a third farm and is planning a fourth. If all are built, they will cover nearly 800 square miles of the North Sea.

And there are plans for many more. In January, the Scottish government’s Crown Estate Scotland agency announced option agreements covering more than 2,800 square miles of sea, with total generating capacity two and half times that of all of the UK’s current offshore turbine fleet.

Growth in wind power looks likely to accelerate further, given soaring gas prices following Russia’s invasion of Ukraine and the UK government’s 8 March decision to phase out imports of Russian oil and gas this year.

UK-headquartered energy group BP, which has extracted oil and gas from the North Sea for more than half a century, was one of the successful ScotWind applicants. It plans to build turbines on a 330 square mile area, starting about 35 miles south-east of Aberdeen with German energy supplier EnBW.

Sophia Fannon-Howell, BP’s global data and analytics principal for sectors including wind, says data played a significant role in shaping the application, including its choice of location and the decision to use turbines fixed to the seabed, drawing on public and in-house data, including from oil and gas projects and shipping.  

But although important for planning wind power, data comes into its own for maintenance. BP gathers real-time data from equipment including oil and gas production facilities and onshore wind turbines, recording more than a billion events each day.

“It’s so critical for us to use analytics in many ways, to run our operations as safely as we can and at the lowest cost that we can,” says Fannon-Howell. “To get the world to net zero quickly, we need to leverage all the data we can to drive down the cost of renewables.”

BP is rapidly expanding renewable energy production as part of its ambition to reduce its corporate carbon emissions to net zero – with the effect of remaining emissions cancelled out by offsetting – by 2050.

“It’s so critical for us to use analytics in many ways, to run our operations as safely as we can and at the lowest cost that we can”
Sophia Fannon-Howell, BP

The company expects to draw on its oil and gas experience in setting up predictive maintenance for offshore wind turbines, including use of robots to gather data on the structural integrity of marine equipment and optimising maintenance logistics. It also owns Onyx Insight, a UK-based developer of condition monitoring systems (CMS) for wind turbines, through its BP Launchpad business accelerator.

“Sometimes the answer isn’t to do everything ourselves,” says Fannon-Howell. BP has installed Onyx’s ecoCMS equipment on 585 turbines in 10 onshore wind farms in the US, and the company reckons this saves nearly $4,800 (£3,600) per turbine each year.

Onyx, which helps to manage about 9,000 wind turbines in 30 countries, gathers a wide range of operational data with ecoCMS, which is analysed through predictive analytics that uses machine learning. It tracks vibrations from the nacelle – the unit that houses the generating equipment to which the rotor and blades are attached – as well as oil levels and data from third-party sensors.

“It’s all around planning maintenance more effectively,” says John Coultate, head of product development. “We detect failures in components more than 12 months in advance and we can monitor that failure progressing.”

Unplanned failures mean losing energy worth thousands of pounds a day while parts are ordered, delivered and installed, in some cases taking months. They can also cause wider damage – a failed bearing can wreck an entire gearbox, requiring a crane to replace and costing hundreds of thousands of pounds.

Accurate predictions can also allow engineers to replace a set of soon-to-fail parts in one visit, which is particularly important at sea, where the company says jack-up vessels needed to replace major components can cost about £100,000 a day.

Coultate says most new turbines have built-in CMSs, so the company is adding the ability to monitor more components, last year including the pitch bearings that join the hub to the blades. Some issues can be detected by joining up several data sources, with one customer planning its maintenance of turbine hydraulic systems with a neural net-based analysis of data from a range of components and vibrations.

“That enables us to give them a list of turbines they should be looking at, rather than going to every one every six months,” says Toby Rogers, head of software.

Digital twin models

There are ways to take such monitoring further. Consultancy Capgemini creates digital twin models of turbines for its clients, three-dimensional visualisations that can be used with virtual reality headsets as well as screens. They enable users to virtually break open the turbine to see individual components, explore issues and predict future failures.

“The biggest benefit is not wasting the time of your engineers to go to a site just to do a standard inspection,” says Graham Upton, chief architect of intelligent industry at Capgemini. “You might have a standard maintenance regime that might be too onerous – you’re going in too often and wasting money. This way, it might say you’ve got a developing problem, you need to go and look at it before it fails.”

This can cut at least a quarter off maintenance costs, although some checks, such for corrosion, still have to be done on-site, says Upton.

Capgemini has built digital twin maintenance services for other industries, including for an Airbus A350 wing factory in Spain. Upton says there is also potential for it to be used in decommissioning nuclear power plants and monitoring railway networks, where models of infrastructure such as tunnels could be updated regularly with cameras fitted to trains.

Remote monitoring can be retrofitted to existing turbines, helping them to stay in use. Fine Energy, a Birmingham-based company that operates and maintains turbines, has developed a bolt-on monitoring system, Osmium, with support from the UK government’s Digital Catapult innovation agency. This resulted from the company taking on maintenance of 14 turbines in Scotland and Italy that lacked remote monitoring options, leading it to develop its own Linux-based hardware to do the job.

“The guiding principle is, we wanted it to be like someone was there,” says Fine Energy director Graham Hygate, adding that remote monitoring needed to be financially viable. “There is no way this fleet of turbines would have carried on operating without this,” he says.

Read more about renewable energy and IT

Osmium develops a digital twin of its turbine within its on-site computer rather than centrally, although some operational changes can be made remotely. For example, the company used it to decrease the cut-off wind speed for one turbine from 25 metres a second to 15, having seen it operates in a very turbulent area. “It’s not really wind speed that wears out wind turbines – they like it,” says Hygate. “It’s turbulence.”

In other situations, remote monitoring means that an engineer can fit replacement parts in one visit, rather than requiring an initial inspection to work out what needs changing.

The system can also be used to improve turbine performance. One Italian turbine operated by Fine Energy was generating less electricity than its wind speed curve predicted. By applying machine learning to the data gathered, the company was able to prove a hunch that dips in performance matched changes in wind direction.

The turbine’s yaw motion – its ability to rotate around its vertical axis to point the nacelle and blades directly into the wind – was hindered by problems with the hydraulics that carry out the turning. “This illustrates the importance of having quite rich information,” says Hygate.

“It’s a way of getting the benefits of Industry 4.0 without having to spend a lot of money on new equipment”
Graham Hygate, Fine Energy

The system is being trialled in other industries, including food processing and hydraulic oil filtration, as a way to add monitoring to existing machinery: “It’s a way of getting the benefits of Industry 4.0 without having to spend a lot of money on new equipment,” he adds.

Octopus Energy Group, a UK-based renewable generator and supplier, includes equipment monitoring within its KrakenFlex management software suite. It gathers a range of real-time data from wind turbines, including operational settings of components including valves, fans and pumps to check for effectiveness and possible faults, and can also use vibration data to monitor components such as blades, drivetrains and gearboxes.

Devrim Celal, chief executive of KrakenFlex, says that using a single software platform helps in making holistic decisions about the group’s operations. “For example, we can tell a battery to send energy back to the grid during peak periods when people are using more energy, and charge the battery up when there is an abundance of cheap, green energy on the grid,” he says.

“We can also make solar and wind farms work more for our customers by increasing their usage when the sun is shining and wind is blowing more than forecast.” This helps wind turbines to generate extra income by allowing them to provide grid balancing services, which require technical capabilities for reporting, control and integration with trading systems.

The software does not yet manage offshore wind farms, says Celal, “but the value of predictive technology and data collection could be even more important for such assets due to the logistical costs in repairing and performing ongoing maintenance of offshore wind farms, especially floating ones”.

Specific data requirements

Given their environment, offshore wind farms have some specific data requirements. Vaarst, a Bristol-based marine robotics specialist, provides an underwater measurement system that builds three-dimensional models that can be accessed remotely, reducing the number of people who need to physically visit and avoiding the safety risks involved with using divers.

Robotic measurements can be used to survey the seabed and look for unexploded ordnance, as well as to check grid connection cables, which can be damaged by abrasion, and mooring cables used for floating platforms.

With rapidly growing numbers, offshore wind turbines will increasingly affect each other’s performance. They remove energy from the wind, creating a wake – a slower wind speed – downstream, an effect more pronounced at sea than on land.

Bronwyn Sutton, principal for offshore wind at Vancouver-based energy performance software provider Clir Renewables, says this means it can make sense to point turbines on the windward edge of a farm slightly out of the wind. This makes them less effective, but improves the performance of those behind and can lead to higher output overall. However, this requires wind farms to be managed holistically rather than individually.

Floating platforms, which allow turbines to be deployed in new offshore locations and will be used in 10 of the 17 ScotWind projects, make things even more complicated. Sutton says there are only a handful of small sites currently in operation and that the floating platforms have their own supervisory control and data acquisition (Scada) systems to manage ballast shifting.

“We’re just at the beginning of understanding these,” she says. To allow this technology to work effectively, the data and software used in wind power will need to become even more powerful.

Read more on Big data analytics

Data Center
Data Management